Bonnie Codes

AI coding agents that understand your project's architecture before they make changes.

Quick start

Install, authenticate through Bonnie, and index your project. Then use any Bonnie agent with architecture-aware context.

npm install -g @bonniebuilds/codes
bonnie login

Navigate to a project and index it:

cd /path/to/your/project
bonnie index

Client setup

Bonnie Codes works with OpenCode, Paseo, VS Code, Cursor, Claude Code, and MCP-compatible clients.

OpenCode

Registers Bonnie as an MCP server and adds Bonnie agents. Restart OpenCode after setup.

bonnie setup opencode

Paseo

Configures Bonnie as an ACP provider. Restart Paseo and select "Bonnie Codes".

bonnie setup paseo

VS Code

Registers Bonnie's MCP server through the VS Code extension.

bonnie setup vscode

Authentication

bonnie login opens a browser for sign-in. Paste the generated token back into the terminal.

bonnie login

For headless environments, pass a token directly:

bonnie login --token bon_<your-token>

Memory

Cloud memory is the launch default after login. Run bonnie index from a repo and Bonnie stores searchable architecture memory for that repo.

  • Free includes one active repo with cloud memory for that repo.
  • Pro monthly and Pro yearly include cloud memory for additional repos.
  • Memory service tokens use opaque mst_... credentials and are separate from CLI bon_... auth tokens.

Hosted models

Hosted Bonnie inference is available on Pro. Bonnie routes hosted model calls through Bonnie-managed Qwen and OpenRouter provider paths and records Bonnie usage/cost receipts on your Bonnie Codes account.

  • Free users do not receive hosted model tokens.
  • Pro monthly and Pro yearly users can use hosted Bonnie inference.
  • Hosted provider billing is managed by Bonnie; users do not manage direct Qwen or OpenRouter billing for hosted inference.

BYO providers

Free and Pro users can configure local-only BYO model providers in their own environment. BYO provider keys stay on your machine and are not stored in Bonnie cloud, Convex, memory artifacts, or hosted gateway config.

CLI bon_... tokens authenticate your Bonnie account only. They are not memory tokens, hosted gateway bearers, or provider API keys.

Agents

Bonnie provides AI agents for development workflows. Each agent loads relevant instructions automatically.

Primary agents

Design

Full design-to-plan workflow: intake, design authoring, review, and implementation planning.

Domain Creator

End-to-end creation of new product domains from requirements to execution.

Debug

Memory-backed debugging with structured root cause classification.

Phase Executor

Orchestrates multi-phase implementation plans with specialized subagents.

Subagents

Design Author

Writes design docs from rough input

Design to Plan

Designs to phased implementation plans

Auditer

Reviews designs, plans, and code

Scout

Fast evidence gathering

Phase Intake

Scope discovery and requirements

Phase Implementer

Code within phase scope

Phase Verifier

Verification checks

Phase Closer

Quality gate and scoring

Phase Committer

Stage and commit phases

Plan

Architecture analysis

Build

Implementation and refactoring

CLI reference

bonnie login                  Authenticate with Bonnie cloud
bonnie login --token <t>     Authenticate with paste-in
bonnie init                   Initialize a project
bonnie index                  Index a project

bonnie setup opencode         Configure OpenCode MCP
bonnie setup paseo            Configure Paseo ACP
bonnie setup vscode           Configure VS Code extension

Account and billing

Use CLI login to create the one-time CLI token handoff for bonnie login. Use Bonnie Codes billing to review your account, start Pro checkout, or manage billing through your BetterAuth-authenticated Bonnie account.

Web billing is tied to your signed-in Bonnie account, not to a pasted CLI token.

Pricing

Free

$0

  • ✓ One active repo
  • ✓ Cloud memory for that repo
  • ✓ Local-only BYO model support
  • ✓ Hosted models require Pro

Pro monthly

$15/month

  • ✓ Additional repos
  • ✓ Cloud memory
  • ✓ Hosted Bonnie inference
  • ✓ Bonnie-managed Qwen and OpenRouter

Pro yearly

$200/year

  • ✓ Annual Pro access
  • ✓ Additional repos
  • ✓ Hosted Bonnie inference
  • ✓ Bonnie-billed account management